A custom design method of ergonomic products based on neighborhood topology reconstruction (NTR) was proposed to improve the design efficiency and comfort level of ergonomic products. The 3D reconstruction method was based on the neighborhood topological relation of medical images, the ambiguity of Marching Cubes algorithm was overcame, and the time-consuming problem of Marching Tetrahedrons algorithm was avoided. The original shape of complex curved surface component with personalized customization information was obtained by 3D reconstruction based on medical CT images, which provided data support for the custom design of ergonomic products. A deep residual network was introduced, and the multi-scale features of the layer cross-section were extracted layer by layer by using the neural network. The nonlinear implicit relationship between cost consumption of additive manufacture and multi-scale features was established layer by layer. The materials consumption prediction and cost optimization of the complex conceptual design prototype were realized. According to the original shape of the manifold and deformation algorithm based on the Laplace-Gauss curve, the hand pressing-holding posture was obtained. The scheme of ordinary mouse was evolved according to the posture, and the ergonomic mouse was designed conceptually. The effectiveness of the method was verified by physical experiments. The high surface precision of the prototype product was indicated by the microscopic morphology, and the predicted energy consumption change was similar to the actual energy consumption. The experimental results show that the combination of neighborhood topology reconstruction and deformation algorithm can provide data support and physical reference for the custom design of ergonomic products and improve the comfort level of ergonomic products.